Why Healthcare AI Needs Clinicians in the Room
This article discusses the importance of involving clinicians in the development of healthcare AI systems to ensure they are practical and effective in real-world settings.
Why it matters
Involving clinicians in the development of healthcare AI is crucial to ensure the models are practical, effective, and trustworthy in real-world clinical settings.
Key Points
- 1Healthcare AI is often built by technologists without clinical experience, leading to models that perform well on paper but struggle in practice.
- 2The author's approach focuses on using only admission-available variables, ensuring interpretability, addressing fairness without removing features, and prioritizing clinical utility over accuracy metrics.
- 3The author is building tools to support emergency care teams with early identification of high-risk patients, resource allocation insights, transparent explanations, and equity-aware predictions.
Details
The author, who started in pharmacy before transitioning to data science, emphasizes the importance of involving clinicians in the development of healthcare AI. Most healthcare AI is built by technologists who lack experience navigating the realities of clinical workflows and patient care delivery. As a result, the models may perform well on paper but fail to be practical or effective in real-world settings. The author's approach focuses on using only variables available within 90 minutes of patient arrival, ensuring the interpretability of the model's predictions, addressing fairness without removing features that may proxy for systemic inequities, and prioritizing clinical utility over pure accuracy metrics. The author is building tools to support emergency care teams, including early identification of high-risk patients, resource allocation insights, transparent explanations to build clinician trust, and equity-aware predictions to advance health justice.
No comments yet
Be the first to comment